CAXF-LCCDE: An Enhanced Feature Extraction and Ensemble Learning Model for XSS Detection

Abstract

Threats to businesses, organizations, and individuals have become more complicated and varied due to the constant evolution of cyberattack strategies brought about by the quick development of Internet technologies. Cross-Site Scripting (XSS) remains one of the most pervasive and dangerous threats among the many that target online application security. This emphasizes the importance of strong detection mechanisms to protect user information and preserve system integrity. Combining machine learning with intrusion detection systems successfully addresses this problem and enhances XSS detection capabilities. This work provides a new feature extraction technique designed specifically for XSS, called Comprehensive Analysis of XSS Features (CAXF). We combine it with Leader Class and Confidence Decision Ensemble (LCCDE), an ensemble machine learning approach. This combination efficiently identifies various XSS attack types, such as reflected, stored, and DOM-based variations. Extensive trials were carried out on well-annotated datasets with typical samples of these XSS categories. The findings demonstrate that the suggested CAXF-LCCDE framework significantly outperforms conventional resilience and detection accuracy models. Furthermore, the results highlight how crucial tailored feature engineering and algorithm selection are for cybersecurity applications. This work provides important information for creating stronger and more dependable XSS attack defenses.